Estimating of Birth Weight Using Placental Characteristics in The Presence of Multicollinearity

Abstract

In this study, it was aimed to compare the performance of proposed estimators in the presence of multicollinearity that will be used in regression analysis as an alternative to Least Squares. Birth weight was estimated by using placental features such as sex, placental efficiency, total cotyledon numbers, large cotyledon weight, medium cotyledon weight, small cotyledon weight, large cotyledon number, medium cotyledon number, small cotyledon number, large cotyledon width, medium cotyledon width, small cotyledon width, large cotyledon length, medium cotyledon length, small cotyledon length, large cotyledon depth, medium cotyledon depth, small cotyledon depth for Bafra sheep breed. In the presence of multicollinearity, more reliable models can be obtained by using some estimator. The performances of the Ridge and Liu estimators, which are suggested methods for this situation, were compared. MSE, RMSE, rRMSE, MAPE, R 2 , and AIC were used as model comparison criteria. As a result of, in the presence of multicollinearity; Liu estimator is recommended as an alternative method to Least Squares.

Comparison of the data mining and machine learning algorithms for predicting the final body weight for Romane sheep breed

“…In literature, there are many studies in which multivariate statistical methods are used together with biometric measurements to create a characterization for breeds [16][17][18][19]. Within the framework of the regression analysis assumptions, requirements such as linearity, normal distribution, constant variance, and explanatory variable independence should be sought in multivariate statistical approaches [20][21][22][23]. However, various data mining and machine learning algorithms do not need these assumptions [24,25].…”